Crack Healing in Mullite-Based EBC during Thermal Shock Cycle

: Crack healing phenomena were observed in mullite and mullite + Yb 2 SiO 5 environmental barrier coating (EBC) materials during thermal shock cycles. Air plasma spray coating was used to deposit the EBC materials onto a Si bondcoat on a SiC f / SiC composite substrate. This study reveals that unidirectional vertical cracks (mud cracks) formed after several thermal shock cycles; however, the cracks were stable for 5000 thermal shock cycles at a maximum temperature of 1350 ◦ C. Moreover, the crack densities decreased with an increasing number of thermal shock cycles. After 3000 thermal shock cycles, cracks were healed via melting of a phase containing SiO 2 phase, which partially ﬁlled the gaps of the cracks and resulted in the precipitation of crystalline Al 2 O 3 in the mullite. Post-indentation tests after thermal shock cycling indicated that the mullite-based EBC maintained its initial mechanical behavior compared to Y 2 SiO 5 . The indentation load–displacement tests revealed that, among the materials investigated in the present study, the mullite + Yb 2 SiO 5 EBC demonstrated the best durability during repetitive thermal shocks.


Introduction
Gas turbines that operate at high temperatures are being developed for and installed in stationary and aviation applications. Higher operating temperatures can increase the energy efficiency and contribute to a cleaner environment. However, because materials that can withstand temperatures greater than 1000 • C are required, studies to develop a ceramic-based material beyond a single-crystalline superalloy are needed. In particular, all-ceramic materials are likely to be used for hot gas components such as combustor liners, blades, and nozzles, which require durable materials that can withstand high temperatures.
Silicon carbide exhibits both excellent heat resistance and high strength at high temperatures in addition to exceptional thermal shock resistance [1,2]. Silicon carbide also exhibits high hardness at room temperature and excellent wear resistance. The problem of brittleness can be overcome through the use of silicon carbide fibers (i.e., SiC-fiber-reinforced SiC (SiC f /SiC) composites). Even if the silicon carbide matrix cracks, the fibers with relatively high mechanical strength support the mechanical load and suppress crack propagation, thereby preventing complete failure [3,4]. SiC f /SiC composites are also expected to resist high-temperature creep and fatigue.
However, high-temperature oxidation and hot corrosion remain unsolved problems impeding the use of SiC f /SiC composite materials in gas turbines. Because gas turbines operate in high-temperature environments for extended periods, the oxide film initially formed on their surface by the oxidation of silicon carbide is expected to function as a protective film that prevents further oxidation. However,

Thermal/Mechanical Characterization
The surface of the coating layer was ground and then polished with 15, 6, 3 and 1 µm diamond suspensions to observe the cracks in the EBCs. The thickness was measured at each polishing step so that the degree of polishing was controlled to be the same for all samples.
Thermal shock tests were performed to evaluate the durability of the EBC against repetitive thermal fatigue. The thermal shock was induced using a laboratory-made horizontal thermal shock machine in which the chamber could be moved horizontally by a motor. The schematic diagram of the thermal shock test is shown in Figure 1a. The test samples were placed on a fixed hotplate such that the substrate was positioned to receive heat from the lower part of the hot plate, and the coating layer was exposed to the top to receive heat from the chamber. First, the test samples placed on the hotplate were maintained at a temperature of 1100 and 1350 • C in the upper coating layer after the heating chamber was closed; the heating rate was 5 • C/min. After the target temperature had been maintained for 10 min, the upper chamber was moved to expose the sample to room temperature for 4 min as shown in Figure 1b, and then the chamber was moved back to expose the sample at 1350 • C again. This cycle was defined as one cycle, and repetitive thermal shocks were applied for a maximum of 5000 cycles.
Coatings 2019, 9, x FOR PEER REVIEW 4 of 14 machine in which the chamber could be moved horizontally by a motor. The schematic diagram of the thermal shock test is shown in Figure 1a. The test samples were placed on a fixed hotplate such that the substrate was positioned to receive heat from the lower part of the hot plate, and the coating layer was exposed to the top to receive heat from the chamber. First, the test samples placed on the hotplate were maintained at a temperature of 1100 and 1350 °C in the upper coating layer after the heating chamber was closed; the heating rate was 5 °C/min. After the target temperature had been maintained for 10 min, the upper chamber was moved to expose the sample to room temperature for 4 min as shown in Figure 1b, and then the chamber was moved back to expose the sample at 1350 °C again. This cycle was defined as one cycle, and repetitive thermal shocks were applied for a maximum of 5000 cycles. We observed the surface using an optical microscope (Olympus, GX51, Tokyo, Japan) after thermal shock tests of 1000, 2000, 3000, 4000, and 5000 cycles. Cracks and damages were observed with an optical microscope, and the density of cracks was measured with an image analyzer. For this purpose, the surface was observed at 200× magnification using an optical microscope. The crack density was defined by measuring the number of cracks intercepting the baseline after the baseline had been marked at positions of 1/4, 1/2, and 3/4 of each line of the horizontal, vertical, and diagonal length, respectively. We defined the crack density as the number of cracks divided by the length of the baseline and expressed it in terms of the number per inch. The numbers of cracks were determined from 10 pictures, and the mean values were plotted. The crack morphology of the surface was also observed with a high-resolution optical microscope (VHX-5000, Keyence Co., Itasca, IL, USA) after thermal shock tests of 100 and 5000 cycles.
Scanning electron microscopy (SEM) was used to observe the cracks in detail after crack healing. The electron microscope (JSM-6010F, JEOL Co., Tokyo, Japan) was operated at an acceleration voltage of 10 kV. Energy-dispersive X-ray spectroscopy (EDS) was used to determine the element species with an acceleration voltage of 10 kV and a spot size of 3.0.
The phase of the coating layer was analyzed by X-ray diffraction (XRD, Rigaku Co., Ltd., Tokyo, Japan) to analyze the change in the phase before and after the crack healing. The XRD analyses were conducted with Cu Kα radiation generated at 45 kV and 200 mA with a scan rate of 2°/min over the 2θ range from 20° to 80°.
Spherical indentations were induced to compare and evaluate the mechanical behavior before and after the thermal shock test. The schematic diagram of spherical indentation is shown in Figure  2. Indentation load-displacement curves were obtained during indentation tests. A tungsten carbide We observed the surface using an optical microscope (Olympus, GX51, Tokyo, Japan) after thermal shock tests of 1000, 2000, 3000, 4000, and 5000 cycles. Cracks and damages were observed with an optical microscope, and the density of cracks was measured with an image analyzer. For this purpose, the surface was observed at 200× magnification using an optical microscope. The crack density was defined by measuring the number of cracks intercepting the baseline after the baseline had been marked at positions of 1/4, 1/2, and 3/4 of each line of the horizontal, vertical, and diagonal length, respectively. We defined the crack density as the number of cracks divided by the length of the baseline and expressed it in terms of the number per inch. The numbers of cracks were determined from 10 pictures, and the mean values were plotted. The crack morphology of the surface was also observed with a high-resolution optical microscope (VHX-5000, Keyence Co., Itasca, IL, USA) after thermal shock tests of 100 and 5000 cycles.
Scanning electron microscopy (SEM) was used to observe the cracks in detail after crack healing. The electron microscope (JSM-6010F, JEOL Co., Tokyo, Japan) was operated at an acceleration voltage of 10 kV. Energy-dispersive X-ray spectroscopy (EDS) was used to determine the element species with an acceleration voltage of 10 kV and a spot size of 3.0.
The phase of the coating layer was analyzed by X-ray diffraction (XRD, Rigaku Co., Ltd., Tokyo, Japan) to analyze the change in the phase before and after the crack healing. The XRD analyses were conducted with Cu Kα radiation generated at 45 kV and 200 mA with a scan rate of 2 • /min over the 2θ range from 20 • to 80 • .
Spherical indentations were induced to compare and evaluate the mechanical behavior before and after the thermal shock test. The schematic diagram of spherical indentation is shown in Figure 2. Indentation load-displacement curves were obtained during indentation tests. A tungsten carbide ball with a radius of r = 3.18 mm was attached to the jig of a universal testing system (Instron 5567, Boston, MA, USA) and the pre-load was applied to the coating layer to P = 10 N at a rate of 3 N/min. The load was then applied to P = 500 N at a rate of 0.1 mm/min. The load was removed at the same rate as soon as the load had reached P = 500 N. The displacement was measured with an extensometer during load changes. Tests were conducted on more than five areas of the coating layer of the test specimens, and the indentation load and displacement results were averaged and plotted. The relative hardness was calculated from relative displacement after unloading in each indentation load-displacement and compared [23][24][25][26]. during load changes. Tests were conducted on more than five areas of the coating layer of the test specimens, and the indentation load and displacement results were averaged and plotted. The relative hardness was calculated from relative displacement after unloading in each indentation load-displacement and compared [23][24][25][26].

Results and discussion
The cracks formed on the surface and cross section of the EBC coating layer by thermal shock were observed by optical microscopy, as shown in Figure 3. Thermal shock tests were carried out at 1350 °C on the coating surface of the sample and at 1100 °C on the bottom of the sample because a gas turbine operating at 1600 °C is cooled simultaneously by the cooling gas through the flow path. Mud-type cracks formed on the surface. After 100 thermal shock cycles, cracks developed in the EBC layer fabricated by APS, as shown in Figure 3a. We considered that these cracks developed because of the difference in thermal expansion coefficients between the EBC coating and the metal sublayer. Several researchers have observed and reported mud cracks in EBCs [8]. However, cross-sectional observations of these cracks reveal that they formed in a direction perpendicular to the interface, not parallel to the interface. The vertical cracks shown in Figure 3b propagated through 400 m of thickness; their propagation stopped at the top of the bondcoat layer. Such cracks can be stable if they do not induce horizontal cracking along the interface between the EBC and the sublayer, but rather tolerate strain caused by stress, thereby contributing to preventing delamination of the coating layer [27,28].

Results and Discussion
The cracks formed on the surface and cross section of the EBC coating layer by thermal shock were observed by optical microscopy, as shown in Figure 3. Thermal shock tests were carried out at 1350 • C on the coating surface of the sample and at 1100 • C on the bottom of the sample because a gas turbine operating at 1600 • C is cooled simultaneously by the cooling gas through the flow path. Mud-type cracks formed on the surface. After 100 thermal shock cycles, cracks developed in the EBC layer fabricated by APS, as shown in Figure 3a. We considered that these cracks developed because of the difference in thermal expansion coefficients between the EBC coating and the metal sublayer. Several researchers have observed and reported mud cracks in EBCs [8]. However, cross-sectional observations of these cracks reveal that they formed in a direction perpendicular to the interface, not parallel to the interface. The vertical cracks shown in Figure 3b propagated through 400 µm of thickness; their propagation stopped at the top of the bondcoat layer. Such cracks can be stable if they do not induce horizontal cracking along the interface between the EBC and the sublayer, but rather tolerate strain caused by stress, thereby contributing to preventing delamination of the coating layer [27,28].
observations of these cracks reveal that they formed in a direction perpendicular to the interface, not parallel to the interface. The vertical cracks shown in Figure 3b propagated through 400 m of thickness; their propagation stopped at the top of the bondcoat layer. Such cracks can be stable if they do not induce horizontal cracking along the interface between the EBC and the sublayer, but rather tolerate strain caused by stress, thereby contributing to preventing delamination of the coating layer [27,28].  The results of the measurements of the crack density developed on the surface of the EBC by thermal shock are shown in Figure 4. The periodic repetition of heating and cooling in turbine systems, and especially an abrupt stop of the system, will exacerbate crack formation. Therefore, the stability of cracks was evaluated by thermal shock tests. The mullite-based EBCs (i.e., the mullite and mullite + Yb 2 SiO 5 EBCs) were not delaminated when the thermal shock cycle of exposure to room temperature was repeated for 5000 cycles while the coating material was maintained at 1350 • C and the SiC f /SiC substrate material was maintained at 1100 • C. Even though a crack is generated initially, interface delamination does not occur because vertical cracks are formed. That is, an initially developed crack in the EBC layer is very stable. Although the Y 2 SiO 5 EBC was stable for 4000 cycles, interface delamination occurred after 5000 thermal shock cycles. The crack density on the surface of a sample after each 1000 cycles in the thermal shock test was measured and plotted in Figure 4. As shown in the graph, the changes in crack density depend on the EBC material. The crack density did not substantially change compared with the initial crack density in the case of Y 2 SiO 5 , whereas the crack density of the mullite-based EBC increased slightly and then drastically decreased after 3000 cycles. Furthermore, it exhibited a substantial decrease in crack density after 5000 cycles. The results of the measurements of the crack density developed on the surface of the EBC by thermal shock are shown in Figure 4. The periodic repetition of heating and cooling in turbine systems, and especially an abrupt stop of the system, will exacerbate crack formation. Therefore, the stability of cracks was evaluated by thermal shock tests. The mullite-based EBCs (i.e., the mullite and mullite + Yb2SiO5 EBCs) were not delaminated when the thermal shock cycle of exposure to room temperature was repeated for 5000 cycles while the coating material was maintained at 1350 °C and the SiCf/SiC substrate material was maintained at 1100 °C . Even though a crack is generated initially, interface delamination does not occur because vertical cracks are formed. That is, an initially developed crack in the EBC layer is very stable. Although the Y2SiO5 EBC was stable for 4000 cycles, interface delamination occurred after 5000 thermal shock cycles. The crack density on the surface of a sample after each 1000 cycles in the thermal shock test was measured and plotted in Figure 4. As shown in the graph, the changes in crack density depend on the EBC material. The crack density did not substantially change compared with the initial crack density in the case of Y2SiO5, whereas the crack density of the mullite-based EBC increased slightly and then drastically decreased after 3000 cycles. Furthermore, it exhibited a substantial decrease in crack density after 5000 cycles. The optical microscope images in Figure 5 show the surfaces of the mullite, mullite + Yb2SiO5, and Y2SiO5 after 2000 and 4000 cycles of thermal shock, indicating crack healing. Although obvious cracks are observed after 2000 cycles of thermal shock irrespective of the EBC material (Figure 5a-c), surprisingly, the surface cracks were not observed after 4000 cycles of thermal shock (Figure 5d-f). In the case of Y2SiO5, relatively large macrocracks developed and were maintained without the formation of microcracks; thus, the crack density did not substantially change. By contrast, the population of microscale cracks increased at 2000 cycles and new phases covered the surface completely at 4000 cycles in the mullite-based EBC. The optical microscope images in Figure 5 show the surfaces of the mullite, mullite + Yb 2 SiO 5 , and Y 2 SiO 5 after 2000 and 4000 cycles of thermal shock, indicating crack healing. Although obvious cracks are observed after 2000 cycles of thermal shock irrespective of the EBC material (Figure 5a-c), surprisingly, the surface cracks were not observed after 4000 cycles of thermal shock (Figure 5d-f). In the case of Y 2 SiO 5 , relatively large macrocracks developed and were maintained without the formation of microcracks; thus, the crack density did not substantially change. By contrast, the population of microscale cracks increased at 2000 cycles and new phases covered the surface completely at 4000 cycles in the mullite-based EBC. The cracks on the surface of the mullite-based EBC after 5000 cycles of thermal shock were observed with a high-resolution optical microscope, as shown in Figure 6. The optical microscope images of the surfaces of the mullite-based EBC after 100 cycles are also presented in the figure.
Although mud cracks were noticeable after 100 cycles of thermal shock, the crack gaps were clearly filled after 5000 cycles of thermal shock.  Figure 7 shows the cracks of the mullite-based EBC after 100 and 5000 thermal shock cycles, as observed at high magnification by SEM. Only cracks and grains are observed after 100 cycles of thermal shock; cracks are observed around the grain. By contrast, after 5000 thermal shock cycles, new phases are formed to fill the gaps and cover the cracks. The results of the EDS composition analysis of these phases indicate that they contain only Si and Al, which are the constituents of mullite. We therefore speculate that the mullite itself underwent a change when repeatedly exposed to high temperatures. The cracks on the surface of the mullite-based EBC after 5000 cycles of thermal shock were observed with a high-resolution optical microscope, as shown in Figure 6. The optical microscope images of the surfaces of the mullite-based EBC after 100 cycles are also presented in the figure.
Although mud cracks were noticeable after 100 cycles of thermal shock, the crack gaps were clearly filled after 5000 cycles of thermal shock. The cracks on the surface of the mullite-based EBC after 5000 cycles of thermal shock were observed with a high-resolution optical microscope, as shown in Figure 6. The optical microscope images of the surfaces of the mullite-based EBC after 100 cycles are also presented in the figure.
Although mud cracks were noticeable after 100 cycles of thermal shock, the crack gaps were clearly filled after 5000 cycles of thermal shock.  Figure 7 shows the cracks of the mullite-based EBC after 100 and 5000 thermal shock cycles, as observed at high magnification by SEM. Only cracks and grains are observed after 100 cycles of thermal shock; cracks are observed around the grain. By contrast, after 5000 thermal shock cycles, new phases are formed to fill the gaps and cover the cracks. The results of the EDS composition analysis of these phases indicate that they contain only Si and Al, which are the constituents of mullite. We therefore speculate that the mullite itself underwent a change when repeatedly exposed to high temperatures.  Figure 7 shows the cracks of the mullite-based EBC after 100 and 5000 thermal shock cycles, as observed at high magnification by SEM. Only cracks and grains are observed after 100 cycles of thermal shock; cracks are observed around the grain. By contrast, after 5000 thermal shock cycles, new phases are formed to fill the gaps and cover the cracks. The results of the EDS composition analysis of these phases indicate that they contain only Si and Al, which are the constituents of mullite. We therefore speculate that the mullite itself underwent a change when repeatedly exposed to high temperatures.  Figure 8 shows the cracks on the cross-section of the mullite, mullite + Yb2SiO5, and Y2SiO5 after 5000 thermal shock cycles, as observed by SEM. Each arrow in the mullite, mullite + Yb2SiO5 indicate crack healed region. On the other hand, each arrow in the Y2SiO5 indicate cracked region. Only cracks are observed after 5000 cycles of thermal shock in the case of Y2SiO5 (Figure 8c). By contrast, after 5000 thermal shock cycles, new phases are formed to fill the gaps and cover the cracks, and this result is more pronounced in the mullite + Yb2SiO5 (Figure 8b). XRD analysis was performed on the EBC materials after 5000 thermal shock cycles for the analysis of the phases formed around the cracks, as shown in Figure 9. In the case of the mullite material, an Al2O3 phase was detected in addition to the mullite phase. Therefore, the newly formed phases in Figures 7 and 8 are attributed to mullite that melted under repeated thermal shock cycles at high temperature. We speculate that the low-melting-point phases containing SiO2 migrated into the cracks, filling the gaps of the cracks, and that the Al2O3 phases precipitated as crystalline phase during cooling. In particular, Al2O3 phases are considered desirable because they exhibit excellent oxidation and corrosion resistance at high temperatures and can therefore resist oxidation and corrosion of EBC materials after crack healing. Similarly, an Al2O3 phase was detected along with the mullite, Yb2SiO5, and Yb2Si2O7 phases in the case of the mullite + Yb2SiO5 material, as shown in Figure  9b. However, in the case of Y2SiO5, Y2O3, and SiO2 phases were detected along with Y2SiO5 and Y2Si2O7 phases, as shown in Figure 9c. The Y2SiO5 phase melts at high temperatures and is decomposed into crystalline Y2O3 and SiO2, similar to the process that occurs in the mullite-based EBC.   Figure 8 shows the cracks on the cross-section of the mullite, mullite + Yb2SiO5, and Y2SiO5 after 5000 thermal shock cycles, as observed by SEM. Each arrow in the mullite, mullite + Yb2SiO5 indicate crack healed region. On the other hand, each arrow in the Y2SiO5 indicate cracked region. Only cracks are observed after 5000 cycles of thermal shock in the case of Y2SiO5 (Figure 8c). By contrast, after 5000 thermal shock cycles, new phases are formed to fill the gaps and cover the cracks, and this result is more pronounced in the mullite + Yb2SiO5 (Figure 8b). XRD analysis was performed on the EBC materials after 5000 thermal shock cycles for the analysis of the phases formed around the cracks, as shown in Figure 9. In the case of the mullite material, an Al2O3 phase was detected in addition to the mullite phase. Therefore, the newly formed phases in Figures 7 and 8 are attributed to mullite that melted under repeated thermal shock cycles at high temperature. We speculate that the low-melting-point phases containing SiO2 migrated into the cracks, filling the gaps of the cracks, and that the Al2O3 phases precipitated as crystalline phase during cooling. In particular, Al2O3 phases are considered desirable because they exhibit excellent oxidation and corrosion resistance at high temperatures and can therefore resist oxidation and corrosion of EBC materials after crack healing. Similarly, an Al2O3 phase was detected along with the mullite, Yb2SiO5, and Yb2Si2O7 phases in the case of the mullite + Yb2SiO5 material, as shown in Figure  9b. However, in the case of Y2SiO5, Y2O3, and SiO2 phases were detected along with Y2SiO5 and Y2Si2O7 phases, as shown in Figure 9c. The Y2SiO5 phase melts at high temperatures and is decomposed into crystalline Y2O3 and SiO2, similar to the process that occurs in the mullite-based EBC. XRD analysis was performed on the EBC materials after 5000 thermal shock cycles for the analysis of the phases formed around the cracks, as shown in Figure 9. In the case of the mullite material, an Al 2 O 3 phase was detected in addition to the mullite phase. Therefore, the newly formed phases in Figures 7 and 8 are attributed to mullite that melted under repeated thermal shock cycles at high temperature. We speculate that the low-melting-point phases containing SiO 2 migrated into the cracks, filling the gaps of the cracks, and that the Al 2 O 3 phases precipitated as crystalline phase during cooling. In particular, Al 2 O 3 phases are considered desirable because they exhibit excellent oxidation and corrosion resistance at high temperatures and can therefore resist oxidation and corrosion of EBC materials after crack healing. Similarly, an Al 2 O 3 phase was detected along with the mullite, Yb 2 SiO 5 , and Yb 2 Si 2 O 7 phases in the case of the mullite + Yb 2 SiO 5 material, as shown in Figure 9b. However, in the case of Y 2 SiO 5 , Y 2 O 3 , and SiO 2 phases were detected along with Y 2 SiO 5 and Y 2 Si 2 O 7 phases, as shown in Figure 9c. The Y 2 SiO 5 phase melts at high temperatures and is decomposed into crystalline Y 2 O 3 and SiO 2 , similar to the process that occurs in the mullite-based EBC. Rare-earth silicate materials such as Y2SiO5 and Yb2SiO5 have been reported to exhibit less mass loss than mullite in a steam atmosphere at relatively higher temperatures [11,17,29,30]. However, if cracks occur in the coating layer because of thermal shock, the crack healing ability of Y2SiO5 and Yb2SiO5 is considered to be relatively poor. Therefore, studies are needed to develop a eutectic phase or new crack healing agent. The mullite + Yb2SiO5 material suggested in this study exhibits high water vapor resistance because of the Yb2SiO5 phase and exhibits crack healing behavior because of the mullite phase. Therefore, the mullite + Yb2SiO5 can be used as a high-temperature EBC material with crack healing ability.
The surfaces of the EBC materials were contacted with a WC ball with a radius of 3.18 mm to a load P of 500 N and then unloaded to determine the mechanical behavior after crack healing, as shown in Figure 10. The graphs show the test results for the sample before the thermal shock and the sample after 5000 thermal shock cycles. The Y2SiO5 EBC shows a large change from elastic to plastic after 5000 thermal shock cycles, as shown in Figure 10c because the test was carried out on the delaminated coating samples. By contrast, the load-displacement behavior of the mullite-based EBCs, which did not exhibit interfacial delamination even after 5000 thermal shock cycles, did not substantially differ from their initial indentation load-displacement behavior, as shown in Figure  10a,b. In the case of mullite, the residual displacement related to the hardness of the material was slightly increased after thermal shock, and the tangential slope of the stress-strain at unloading, which is related to the modulus of elasticity, decreased after thermal shock [10,25,26,31]. Thus, the hardness and elastic modulus of the sample decreased by 5000 thermal shock cycles. The decrease in the mechanical properties is attributed to cracks even though the sintering of the material during repetitive exposure to high temperatures occurs [32,33]. However, the hardness and elastic modulus Rare-earth silicate materials such as Y 2 SiO 5 and Yb 2 SiO 5 have been reported to exhibit less mass loss than mullite in a steam atmosphere at relatively higher temperatures [11,17,29,30]. However, if cracks occur in the coating layer because of thermal shock, the crack healing ability of Y 2 SiO 5 and Yb 2 SiO 5 is considered to be relatively poor. Therefore, studies are needed to develop a eutectic phase or new crack healing agent. The mullite + Yb 2 SiO 5 material suggested in this study exhibits high water vapor resistance because of the Yb 2 SiO 5 phase and exhibits crack healing behavior because of the mullite phase. Therefore, the mullite + Yb 2 SiO 5 can be used as a high-temperature EBC material with crack healing ability.
The surfaces of the EBC materials were contacted with a WC ball with a radius of 3.18 mm to a load P of 500 N and then unloaded to determine the mechanical behavior after crack healing, as shown in Figure 10. The graphs show the test results for the sample before the thermal shock and the sample after 5000 thermal shock cycles. The Y 2 SiO 5 EBC shows a large change from elastic to plastic after 5000 thermal shock cycles, as shown in Figure 10c because the test was carried out on the delaminated coating samples. By contrast, the load-displacement behavior of the mullite-based EBCs, which did not exhibit interfacial delamination even after 5000 thermal shock cycles, did not substantially differ from their initial indentation load-displacement behavior, as shown in Figure 10a,b. In the case of mullite, the residual displacement related to the hardness of the material was slightly increased after thermal shock, and the tangential slope of the stress-strain at unloading, which is related to the modulus of elasticity, decreased after thermal shock [10,25,26,31]. Thus, the hardness and elastic modulus of the sample decreased by 5000 thermal shock cycles. The decrease in the mechanical properties is attributed to cracks even though the sintering of the material during repetitive exposure to high temperatures occurs [32,33]. However, the hardness and elastic modulus of the mullite + Yb 2 SiO 5 did not substantially change even after 5000 thermal shock cycles. Therefore, the initial mechanical properties of the mullite + Yb 2 SiO 5 were recovered by crack healing.
of thermal shock is plotted in percent. As shown in the graph, the mullite + Yb2SiO5 maintains its initial hardness value. These results indicate that the mullite + Yb2SiO5 is a potential EBC material because gas turbine coating materials exposed to repeated thermal cycles must not only not exhibit interfacial delamination due to the thermal shock but also resist impact and wear due to external FOD.   Figure 11 shows the change in the relative hardness of the material, as revealed by analysis of the residual displacement of the indentation load-displacement curves in Figure 10. The hardness value before the thermal shock test is assumed to be 100%, and the hardness value after 5000 cycles of thermal shock is plotted in percent. As shown in the graph, the mullite + Yb 2 SiO 5 maintains its initial hardness value. These results indicate that the mullite + Yb 2 SiO 5 is a potential EBC material because gas turbine coating materials exposed to repeated thermal cycles must not only not exhibit interfacial delamination due to the thermal shock but also resist impact and wear due to external FOD.
(c) Figure 10. Plot of indentation load-displacement curves of (a) mullite, (b) mullite + Yb2SiO5 and (c) Y2SiO5 using a WC ball with a radius of 3.18 mm at load P = 500 N. Figure 11. Hardness changes of EBC materials during thermal shock tests.

Conclusions
The purpose of this study was to introduce an EBC to protect SiCf/SiC composites from hightemperature oxidation and corrosion and to investigate the durability of the coating when exposed to high temperatures and repeated thermal shocks. Crack formation, crack stability, and crack healing were investigated after specimens were subjected to repeated heating to 1350 °C followed by cooling. In addition, spherical post-indentation tests were performed to determine the change in mechanical behavior after thermal shock. As a result, the following conclusions were obtained: (1) EBC layers composed of mullite, mullite + Yb2SiO5, and Y2SiO5 were coated via APS onto a Si bondcoat. The bondcoat was previously coated onto a SiCf/SiC composite. All of the coating Figure 11. Hardness changes of EBC materials during thermal shock tests.

Conclusions
The purpose of this study was to introduce an EBC to protect SiC f /SiC composites from high-temperature oxidation and corrosion and to investigate the durability of the coating when exposed to high temperatures and repeated thermal shocks. Crack formation, crack stability, and crack healing were investigated after specimens were subjected to repeated heating to 1350 • C followed by cooling. In addition, spherical post-indentation tests were performed to determine the change in mechanical behavior after thermal shock. As a result, the following conclusions were obtained: (1) EBC layers composed of mullite, mullite + Yb 2 SiO 5 , and Y 2 SiO 5 were coated via APS onto a Si bondcoat. The bondcoat was previously coated onto a SiC f /SiC composite. All of the coating process conditions were controlled to achieve a uniform thickness of 400 µm. In addition, the prepared EBC coating layers were expected to exhibit less than 10% porosity to prevent water vapor penetration at high temperatures. (2) Mud cracks developed on the surface and unidirectional vertical cracks formed on the cross section after several thermal shock cycles in the EBC layer fabricated by APS. The thermal shock cycles involved repeated exposure to room temperature after the coating material was maintained at 1350 • C and the SiC f /SiC substrate material was maintained at 1100 • C. However, the cracks in the mullite-based EBC were very stable and did not lead to interfacial delamination until 5000 thermal shock cycles were completed. This behavior was attributed to the advantage of strain tolerance due to vertical cracks. The Y 2 SiO 5 coating material also exhibited good stability: interface delamination did not occur until 4000 thermal shock cycles were completed. (3) Crack healing phenomena were observed in the present study when the thermal shock cycles were repeated for 4000-5000 cycles for the mullite-based EBCs, where the temperature of the EBC coating layer was maintained at 1350 • C and the SiC f /SiC composite material was maintained at 1100 • C. The cracks on the cross-section as well as surface were covered with a new phase in the case of the mullite-based EBC. Microscopic observations revealed that a new phase was formed around the gap of the cracks. A crystalline phase of Al 2 O 3 was detected by XRD analysis. We considered that the mullite phase melted at high temperatures and filled the gaps of the cracks and that a crystalline phase of Al 2 O 3 was precipitated during cooling, resulting in crack healing. Crystalline SiO 2 and Y 2 O 3 phases were observed in the case of Y 2 SiO 5 . (4) The mechanical indentation load-displacement behaviors of the EBC samples were characterized using a spherical indentation method after the specimens were subjected to 5000 thermal shock cycles. The mullite-based EBC showed no substantial change in mechanical behavior even after 5000 thermal shock cycles, whereas the Y 2 SiO 5 exhibited a change from elastic to plastic behavior because of interfacial delamination. In particular, the mullite + Yb 2 SiO 5 EBC maintained its initial mechanical behavior even after 5000 thermal shock cycles because of its crack healing behavior.